Understand AI tools, check their output and build practical workflows with original sources and clear examples.
Search this section ↗
96 topics in this section.
Find out whether a ChatGPT upload failed because of a quota, the file itself or your connection. Use a small test before deleting files or retrying repeatedly.
Change ChatGPT’s model-improvement setting, distinguish it from memory and Temporary Chat, and understand what saving or sharing a conversation changes.
Request a ChatGPT data export, download it from the right account, and check the archive before relying on it as a backup or moving to another workspace.
For an image-to-document task, specify what is visible and what must remain unknown; review facts separately from the generated artifact.
Use AI to organise notes into tasks, owners and dates, then trace each item back to the source. Keep suggestions separate from agreed commitments.
Test an AI-generated Excel formula with small known answers, blank cells and copied rows. Catch a wrong calculation even when Excel shows no error.
Separate facts from advice, verify consequential claims against original sources and mark what you cannot confirm.
Separate source content from authorized instructions, inspect tool requests and prevent a retrieved page from gaining account permissions.
Read the model card’s task, licence and evaluation, then check the files and memory estimate. A practical checklist for choosing a usable model.
A free input/output row does not make every feature free or suitable for confidential data. Read the exact model’s tier and terms.
Check commitments, dates, recipients and attachments, then edit the tone and send only after your own final review.
Crop unnecessary content, cover sensitive details with an opaque edit and reopen the exported image to verify what remains visible.
Map each important summary point to its page, inspect exclusions and tables and distinguish missing evidence from a concise explanation.
Create a minimal reproducible description, replace secrets consistently and request diagnosis before accepting a suggested command.
Separate prose from executable text, keep exact identifiers and verify that the translated steps still describe the same action.
Keep the goal, confirmed decisions, current files and test evidence in a concise record before moving to a new conversation.
Verify quantity, price, discounts, tax and currency against the image before importing a total into records.
Use the same task and evidence, score factual support and usefulness separately and inspect the failures behind your preference.
Choose a short series of guides for the task you want to complete.
Keep your context, check the output and protect the tool boundary.